To enhance the security of schools,shopping malls,and other public places,it is important to achieve auto-matic identification of abnormal two-person interaction behaviors,such as stealing,robbing,fighting,and assaulting,in surveillance videos.However,the current behavior recognition method based on joint data in graph creation neglects the two-person interaction information as well as the interaction relationship between the single unnatural connection joints.To address this issue,a two-person interaction behavior recognition model based on the interactive relationship hyper-graph convolution network is proposed to model and identify human interactions.First,the corresponding single hyper-graph and two-person interaction graph are created for the joint-point data of each frame,where the hypergraph makes the information of multiple unnaturally connected nodes interchangeable at the same time,and the interaction graph em-phasizes the interaction strength between nodes.The above-constructed graph models are fed into the spatiotemporal graph convolution to model the spatial and temporal information separately,and lastly,the recognition results are ac-quired by the SoftMax classifier.The benefits of the proposed algorithm framework are that the interactive relationship between two persons,the structural relationship between unnatural connections,and the flexible motion characteristics of limbs are regarded in the graph construction process.Tests on the NTU data set demonstrate that the algorithm at-tains a correct recognition rate of 97.36%.The findings indicate that the network model enhances the ability to represent the characteristics of two-person interaction and has better recognition performance than the current models.
Traditional vehicle trajectory prediction models widely exist the generalization problem towards unknown scenarios. In this paper, we address the generalization via the following ways. A conditional variational autoencoder based on invariance penalty is adopted to predict trajectory endpoints. In addition, we propose a domain division method to enhance the performance of the invariance principle and design the maneuver-related subtasks to reconstruct the consistent features of trajectories. The experiment is carried out on the INTERACTION dataset, which is well employed in the study of trajectory domains. Compared to the SOTA models, the mADE at 3s decreases from 1.16 to 0.53. The ablation study demonstrates the effectiveness of each module in our model. The results show that our method achieves excellence performance when generalized to unknown domains.
Newspaper pages feature a reasonable and well-organized news layout,allowing readers to access the most valuable information and gain pleasure in a short span.However,for typesetters,generating this type of newspaper lay-out is time-consuming.A method for automatically generating digital newspaper layouts is proposed by combining the Bayesian network inference and constrained programming.An inference model for the structure and attribute of digital newspaper layout is established using historical layout data and expertise to endow newly generated layout with a histor-ically specific style.Then,a mixed integer constraint programming model is proposed to compute the layout using the inference results.This approach aims to markedly reduce the solution space of the model and improve the layout quality.In addition,the inference model provides available candidate structures and generates varying results.The programming model also ensures that the layout has excellent alignment performance.A dataset with Chinese newspaper layouts is es-tablished to train and validate the model;it includes detailed news attribute label data.Research results exhibit the ef-fectiveness of the automated layout generation method.
In traditional newspaper printing industry,designers need to manually typeset according to the design rules.The typesetting process is costly,time-consuming and labor-consuming.In order to improve the efficiency of typeset-ting,a method of automatic style generation and fine-tuning based on historical excellent layouts is proposed.In order to learn the style of newspaper typesetting from the data,an electronic newspaper database containing rich parameter in-formation of design elements is created,which can effectively reflect the layout of newspapers.For a given news article,firstly,a probability model is trained according to the historical excellent layout to infer the layout style of the electronic newspaper,and the fixed constraints and user constraints are combined to ensure that the style is effective.At the same time,some aesthetic design principles are quantified to further realize style fine-tuning.Finally,through qualitative and quantitative evaluation,it shows that the new method can generate newspapers that meet the visual aesthetics,hierarchy and readability.This method can provide a reference for intelligent generation of layout design styles.
The ability of intelligent robots to function in complex environments has been a longstanding challenge in the field of robotic applications. Referential expressions are frequently utilized for object positioning, making this method a common approach in robot interactions. However, relying on a single visual modality alone is not adequate for all tasks in real-world scenarios. This study proposes a robot perception system based on the fusion of visual and auditory modalities. The system employs a deep learning algorithm model to realize the visual and auditory perceptions of the robot,and it processes natural language and scene information for visual positioning and collects data from 12 types of sound signals for audio recognition. The experimental results indicate that the system integrated into the UR robot has a strong visual positioning ability and audio prediction, and it has successfully carried out an instruction-based audio-visual operation task. The results confirm that audio-visual data has a higher expressive capability than single-modal data.
